Critical gaps across all dimensions. Tool definitions lack depth and rigor. Duplicate tool 'add' (tools 1-2) indicates poor curation. Descriptions are minimal (10-200 chars) but lack actionable context for LLM selection. Parameter descriptions exist but are generic. No output schemas documented. No error handling guidance. The 'call_sentinel' tool has complex nested parameters (messages array, guardrails object) with only surface-level descriptions, LLMs cannot infer valid structures. No validation rules, constraints, or recovery paths documented.

• Remove duplicate 'add' tool. Keep only one canonical implementation.
• Document output schemas for all tools. E.g., add/multiply/subtract return {"result": integer}; call_sentinel returns {"scores": {"category": float}, "safe": boolean}.
• Expand call_sentinel parameter descriptions with JSON schema examples: 'messages: Array of {"role": "system"|"user"|"assistant", "content": string}. Default: [{"role": "system", "content": "You are an education bot focused on O Level Maths."}]'.
• Add error handling guidance to call_sentinel: 'Returns 400 if text is empty or exceeds 10000 chars. Returns 401 if SENTINEL_API_KEY is invalid. Returns 503 if Sentinel service is unavailable, retry after 30s.'
• Enhance arithmetic tool descriptions to clarify use cases and constraints: 'Adds two integers. Returns result as integer. Fails if inputs exceed ±2^31-1. Use for simple arithmetic; for complex expressions, call multiple times.'